Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
EVALUATIONS OF MICROSIMULATION MODELS: LITERATURE REVIEW 271 after implementation of the law. Food stamp program costs and participation rose independently of program changes because of increasing unemployment and poverty. The national unemployment rate increased from 5.0 percent in December 1978, which was the month that implementation began, to 7.8 percent in May 1980, approximately 1 year after implementation was essentially completed. Two other problems arose for Beebout and Haworth in carrying out their external validation. First, the 1977 Food Stamp Act was not implemented in most states until the first half of 1979, while the simulation estimates were prepared in mid-1978 and assumed full implementation at that time. Second, because no appropriate behavioral equation was available in the model or the literature, an ad hoc approach was used to model the change in participation rates that resulted from the act, which eliminated the requirement that participating households purchase a portion of their food coupons. The impact on participation that would result from not having this purchase requirement for food stamps was not known, but it was presumed to vary with income adjusted for household size and with receipt of public assistance. Analysts from the U.S. Department of Agriculture's Food and Nutrition Service and Mathematica Policy Research, Inc., made assumptions about how much participation rates were likely to increase for subpopulations defined by income source and level of poverty. Their assumptions were influenced by observed participation rates for cash assistance programs and participation rates of food stamp recipients whose purchase requirements were small or zero. Beebout and Haworth summarize MATH's performance by saying that in the aggregate it appears to have underestimated the 1977 act's impact by 1.8 to 12.8 percentage points and that the low end of this range is probably tolerable to policy makers but that the high end is not. BURTLESS (1989) Burtless (1989) compared MATH with TATSIM, which is a variant of MATH developed by the Stanford Research Institute to predict national-level labor supply responses to alternative negative income tax plans. Again, there was some problem in comparability; the MATH estimates were based on the 1975 March CPS while the TATSIM estimates were based on the 1976 Survey of Income and Education, data that are 1 year older. In addition, MATH's results used preliminary labor supply data from the Seattle-Denver negative income tax experiment, while TATSIM used revised information. However, the variation of results seemed to dwarf any of these considerations. Not only were substantial changes in estimates of work effort and the fraction receiving benefits, among others, seen when shifting from MATH to TATSIM, but changes in sign were discovered for earnings and aggregate labor supply among single mothers. Burtless commented that the results raise troubling questions about the